Structural Expectations and the Effectiveness of Government Policy in a Short-Run Macroeconomic Model
Abstract
Over the past few years, expectations, and in particular inflationary expectations, have come to play a central role in macroeconomic theory. In modeling these expectations two alternative procedures have typically been adopted. One approach is to specify them by some autoregressive function of the variable being predicted, so that at any specified time t, say, they can be treated as given, being predetermined by past values of that variable. The most common of such autoregressive procedures is the adaptive expectations hypothesis in which the forecast is adjusted in proportion to the immediate past forecast error. Despite their widespread use in a variety of contexts, these autoregressive hypotheses have periodically come under severe criticism, especially when applied to predicting endogenous variables. The objections have been along the following lines. By forecasting an endogenous variable using past values of that variable alone, one is clearly disregarding a considerable volume of available information relevant to that variable. In particular one is ignoring any knowledge one might have of the economic structure being analyzed, the very purpose of which is to provide predictions of the endogenous variable. Indeed there is no reason for the predictions generated from the autoregressive scheme to be consistent with those implied by the model. Hence it has been argued that if forecasters are aware of the structure of the relevant economic system, the rational way for them to form their expectations is to base them on the predictions of the economic model. Of course this criticism does not apply to predictions of exogenous variables, since by definition these are not explained within the framework of the model. This hypothesis, known as the rational expectations hypothesis, originated with John Muth. Formally it requires the forecaster's predicted value for period t, say, to equal the expected value of that variable as predicted by the system, conditional on all information available at the time the prediction is made (usually time (t 1)). The insistence that expectations be rational is also open to objections. In order for a prediction to equal the corresponding conditional expected value, it is necessary for the economic agents to have perfect knowledge of the complete economic structure, except for the truly random disturbances. This means that they must know the values of all relevant parameters determining the underlying economic relationships, as well as the means of all exogenous variables, including exogenous government policy variables. While one can argue, as I shall, that given the availability of consistently estimated economic models, it may not be too unreasonable to assume that economic forecasters have unbiased estimates of relevant parameters, it is most unlikely that they will have such knowledge of exogenous variables, especially those under government control. Indeed, as I shall show below, in certain cases to be discussed, it is precisely in the government's interests to deliberately misinform the public as to its proposed policies if it wishes them to be effective. As a result of the overwhelming quantity of information it assumes, the use *Professor of economics, Australian National University. An earlier version of this paper was presented to the Workshop in Macroeconomics at the University of Virginia; I wish to thank participants of the workshop for their helpful comments. The exposition of the paper has benefited from the suggestions of the managing editor and an anonymous referee.
- Sources
- openalex